Multi-Turn testing for AI agents: Check every interaction | Giskard Hub v.3.0

A whole-conversation check tells you the agent failed. It doesn't tell you that the tool call at turn one was already wrong. So you read the trace backwards, guessing which step poisoned the seven that followed. Giskard Hub v3 lets you put the check where the failure is. 🎯 What's new: - Checks at intermediate steps — Assert on turn one, turn two, or any step in the trace, not just the final output. - When a step-level check fails, the conversation stops there. - Tool call verification — Confirm your agent called the right tool at the right moment, before the answer is even generated. - Whole-trace checks, still there — Answer relevance and other end-to-end evaluators run across the full conversation exactly as before. 📖 Documentation: https://docs.giskard.ai/hub/ui 🚀 Request a demo: https://www.giskard.ai/contact #AIsecurity #RedTeaming #LLMEvaluation #AIagents